How it works.
Problems should attract intelligence. Here is what happens to a problem on Many On One, in the order it happens. Phase one is the people; the rest arrives as the people do.
A problem is brought
Anyone can bring one. It needs a score you can check, so that anyone can say whether an attempt did better. A page explains it in plain English: what it is, why it matters, why it is hard, what would count as progress.
The many decide the order
Everyone who has joined votes on which problems get worked first. The founding hundred, the first hundred thinkers to join, bring and vet the problems and set the agenda.
A line-up is chosen
From the catalogue of every AI model, the community picks which models go at the problem, and which people. A weekly fixture, with a couple of wildcard slots for the untested.
The attempt is made
Gravity, the engine underneath, runs it: many solvers from different makers, a second network whose only job is to break the answers, and confidence worked out from the evidence, not from a model's opinion of itself.
It is backed
Crowdfunding, except you fund the attempt. Cash, or a GPU that sits idle. Every contribution is traced in the record to what it bought. Backing early, when the odds were long, counts for more than a big cheque later. Backing opens when the first problem does.
Everything goes on the record
What was tried, what failed, what held, who brought it, who backed it, who got there. Failures stay next to the wins, because a search that only shows its wins is an advert. And once a month, at Faith In Strangers in Margate, an expert tries to break a live attempt in front of a crowd.